Artificial neural networks to model earthquake magnitude and the di-rection of its energy propagation: a case study of Indonesia

Indonesia is a region that experiences frequent earthquakes, and therefore is highly prone to earthquake hazards. Elevated seismicity in Indonesia means that building models to understand and predict earthquake characteristics and their associated hazards is important. The goal of this research is to develop a supervised learning artificial neural network (ANN) application that can predict the magnitude of earthquake wave propagation (bar) and the direction of propagation of earthquake wave using some selected earthquake data (Mw 5-8) happened in Indonesia from 1996 to 2019. The data was taken from the United States Geological Survey (USGS) database and the Indonesian Agency for Meteorological, Climatological, and Geophysics (BMKG) database. The earthquake data used in the artificial neural network application consists of a hidden layer with four neurons and seven input neurons that contain earthquake parameters, including longitude, latitude, magnitude, depth, strike, dip, and rake, as well as one output neuron. The magnitude and direction of energy propagation of the earthquakes were successfully predicted using the ANN program. Excellent agreement between the results of ANN and those of Coulomb 3.3 software strongly indicates that the ANN program can be used as an alternative to the existing Coulomb 3.3 software. The ANN model can also be further applied to other earthquake data around the world. The results are expected to contribute to the development of earthquake detection software tools. With artificial intelligence, earthquake prediction software will be more effective and can reduce the risks of failure in predicting the magnitude and direction of earthquake wave propagation.